• Objectives
• Goalie Overview
• Goalie Performance
• Goalie Contracts
• Predicting Salary
• Findings

• How good are goalies?
• Do they work hard for the money?
• How much to pay goalies?

• We analyzed 187 goalies over the past 7 seasons

• How many games do goalies play in general?

• There are 58 goalies who actually played in the the NHL? (>40 games played)

• Where are NHL goalies from?

Demographics

• What do they look like (height & weight)?

Top 10 Averages By Goalies in the Last 6 Seasons

Top Goalies by Season

Worst 10 Goalie Averages in the Last 6 Seasons

Goalies that had a really bad season

How well the first placed team goalie performed

• Red point is league average

How much do goalies get paid and how has that changed over time?

• How are goalie contracts inflating?

Who are the Top 10 paid players in the last 7 years?

Can we predict how much goalies will get paid?

• Build multivariate regression model to predict contract average annual value
• Make predictions and compare to actual contracts

Feature selection

• Way too many predictors

reg_data %>% ncol()

## [1] 60

Calculate information gain for predictors and grab 10 highest predictors in information gain

## # A tibble: 6 x 11
##   play_continued_… x_on_goal unblocked_shot_… x_freeze   min icetime
##              <dbl>     <dbl>            <dbl>    <dbl> <dbl>   <dbl>
## 1                6      21.0               38     4.45    40    2400
## 2              376     788.              1498   179.    1569   94166
## 3              726    1639.              2844   384.    3000  179926
## 4               41      75.0              145    18.1    139    8023
## 5              241     533.               976   125.    1094   65647
## 6              844    1821.              3330   424.    3084  184714
## # … with 5 more variables: low_danger_shots <dbl>,
## #   high_dangerx_goals <dbl>, gs <dbl>, play_continued_outside_zone <dbl>,
## #   aav <dbl>

Build regression model with selected features

## # A tibble: 11 x 5
##    term                        estimate std.error statistic  p.value
##    <chr>                          <dbl>     <dbl>     <dbl>    <dbl>
##  1 (Intercept)                 686545.   110074.      6.24  9.13e-10
##  2 play_continued_in_zone        1228.     2698.      0.455 6.49e- 1
##  3 x_on_goal                     2157.     5600.      0.385 7.00e- 1
##  4 unblocked_shot_attempts      -1198.     1224.     -0.978 3.28e- 1
##  5 x_freeze                     20710.    20922.      0.990 3.23e- 1
##  6 min                          -2924.     5158.     -0.567 5.71e- 1
##  7 icetime                        -22.1      84.8    -0.261 7.95e- 1
##  8 low_danger_shots             -5425.     2578.     -2.10  3.58e- 2
##  9 high_dangerx_goals           -4783.    19989.     -0.239 8.11e- 1
## 10 gs                          351069.    49850.      7.04  5.90e-12
## 11 play_continued_outside_zone   -807.     2854.     -0.283 7.78e- 1

Model evaluation

## # A tibble: 1 x 6
##   r.squared adj.r.squared    sigma statistic  p.value    df
##       <dbl>         <dbl>    <dbl>     <dbl>    <dbl> <int>
## 1     0.556         0.548 1425037.      66.3 9.95e-87    11

Predictions

• Sergei Bobrovsky: got paid $11M AAV when he signed this season

##       1 
## 5294396

• Mikko Koskinen: got paid $4.5M AAV when he signed an extension

##       1 
## 4391092

• Most of the goalies should probably get paid around $5 Million/year
• How much a goalie gets paid is not always an indicator of how good of a player they actually are

• Questions?